This documentation is archived and is not being maintained.

Architecture Diagram: Runtime and Operations

Speech Server 2007

This content is no longer actively maintained. It is provided as is, for anyone who may still be using these technologies, with no warranties or claims of accuracy with regard to the most recent product version or service release.

The following high-level diagram illustrates the Speech Server components and the relationships between them as applied to the run-time stage of application deployment. This diagram assumes that the organization is running a distributed deployment where components such as Speech Server, Microsoft Operations Manager (MOM), and a Web server run on separate computers.


  1. A call comes into a Session Initiation Protocol (SIP) peer from either a Voice over IP (VoIP) client, a SIP client, or a telephony client (through a PSTN), which prompts the SIP peer to create a request.
  2. The SIP peer contains a dial plan, which maps a phone number to an IP address on the??computer running Speech Server. The SIP peer sends a SIP invite to Speech Server. The SIP invite is received by the Telephony Application Proxy (TAP).
  3. TAP sends an HTTP request for the speax file that activates the requested application. TAP also sends a SIP redirect message to the SIP peer. The redirect message contains the location of the requested application.
  4. The redirect message causes the SIP peer to send a SIP invite to ASP.NET through the TelephonySession. The TelephonySession receives the subsequent invite and creates a remote procedure call (RPC) channel between the TelephonySession and Speech Engine Services??(SES). SES??allocates a speech recognizer, a speech synthesizer, and a dual tone multi-frequency (DTMF) processing engine for the call.
  5. The requested application is started and either accepts or rejects the invite. The SIP peer receives and acknowledges this response.
    1. When the call is established, the TelephonySession raises an event to the application, which can then start speaking and listening.
    2. When the caller hangs up, the endpoint client informs the SIP peer that the call is complete.
    3. The SIP peer relays the message to Speech Server, which releases the resources and informs the application that the call is complete.
  6. Throughout the duration of the call, information is written to the event trace log (ETL) file. Administrators can run the command-line utility MssLogToDatabase.exe to write the content of the ETL file to the SQL??Server tuning database. This utility can be run manually, in a batch file, or using the Windows scheduler. For more information, see Import Log Files into the Tuning Database.
  7. Administrators can use??the Speech Server??Management Pack for MOM 2005 to monitor important errors, warnings, and informational events that??Speech Server??writes to the application event log.